fix: report actual file size in scan summary when scanner exits early

modelaudit/core.py

@@ -1435,6 +1435,14 @@ def _scan_file_internal(path: str, config: dict[str, Any] | None = None) -> Scan
             },
         )
 
+    # Ensure bytes_scanned reflects the actual file size even when a scanner
+    # returns early (e.g. missing optional dependency, parse error).  The file
+    # size was already computed above via os.path.getsize and is guaranteed to
+    # be accurate.  Without this fallback the scan summary reports "Size: 0
+    # bytes" for every file whose scanner didn't explicitly set the field.
+    if result.bytes_scanned == 0 and file_size > 0:
+        result.bytes_scanned = file_size
+
     return result
 
 

modelaudit/detectors/suspicious_symbols.py

@@ -170,6 +170,71 @@
     ],  # dill's load helpers can execute arbitrary code when unpickling
     # References to the private dill._dill module are also suspicious
     "dill._dill": "*",
+    # Dynamic resolution / import trampolines
+    "pkgutil": ["resolve_name", "get_importer", "walk_packages"],
+    "zipimport": "*",
+    # uuid — _get_command_stdout/_popen internally call subprocess.Popen
+    "uuid": ["_get_command_stdout", "_popen"],
+    # Network / exfiltration
+    "smtplib": "*",
+    "xmlrpc": "*",
+    "xmlrpc.client": "*",
+    "xmlrpc.server": "*",
+    "poplib": "*",
+    "imaplib": "*",
+    "nntplib": "*",
+    "ssl": "*",
+    "socketserver": "*",
+    "requests": "*",
+    "aiohttp": "*",
+    # Code execution / compilation
+    "codeop": "*",
+    "marshal": ["loads", "load", "dumps", "dump"],
+    "compileall": "*",
+    "py_compile": "*",
+    # FFI / native code
+    "_ctypes": "*",
+    # Profiling / debugging (can execute code)
+    "cProfile": "*",
+    "profile": "*",
+    "pdb": "*",
+    "timeit": ["timeit", "repeat"],
+    "trace": "*",
+    # Operator / functools bypasses
+    "functools": ["reduce", "partial"],
+    "_operator": "*",
+    # Pickle recursion
+    "cloudpickle": "*",
+    "joblib": "*",
+    # Filesystem / shell
+    "filecmp": "*",
+    "distutils": "*",
+    "pydoc": "*",
+    "pexpect": "*",
+    "fileinput": "*",
+    "glob": "*",
+    # Virtual environments / package install
+    "venv": "*",
+    "ensurepip": "*",
+    "pip": "*",
+    # Threading / process / signal
+    "_signal": "*",
+    "threading": "*",
+    "_thread": "*",
+    # Database / archive / other
+    "sqlite3": "*",
+    "_sqlite3": "*",
+    "select": "*",
+    "selectors": "*",
+    "logging": ["config"],
+    "syslog": "*",
+    "tarfile": "*",
+    "zipfile": "*",
+    "shelve": "*",
+    # Documentation / tooling (can execute code)
+    "doctest": "*",
+    "idlelib": "*",
+    "lib2to3": "*",
 }
 
 # Advanced pickle patterns targeting sophisticated exploitation techniques